RegNetY-10B-SEER Model

Introduction

The RegNetY-10B-SEER is a large-scale vision model from Meta, trained on one billion uncurated Instagram images. It is designed primarily for image classification tasks. Hugging Face hosts this model, providing a convenient way to leverage its capabilities for various image-related use cases.

Architecture

RegNetY-10B-SEER is a scaled-up version of the RegNetY architecture. The model architecture is detailed in the paper here. It utilizes a deep convolutional network structure suitable for high-capacity vision tasks.

Training

The model was trained on a massive dataset of one billion Instagram images, which were not curated or labeled specifically for this training. This dataset choice reflects the model's robustness and adaptability to diverse, large-scale image data.

Guide: Running Locally

To run the RegNetY-10B-SEER model locally, follow these steps:

1. Install Dependencies: Ensure you have the transformers, torch, and datasets libraries.

   pip install transformers torch datasets
   

2. Load the Model and Data: Use the Hugging Face Transformers library to load the model and a sample dataset.

   from transformers import AutoFeatureExtractor, RegNetModel
   import torch
   from datasets import load_dataset
   
   dataset = load_dataset("huggingface/cats-image")
   image = dataset["test"]["image"][0]
   
   feature_extractor = AutoFeatureExtractor.from_pretrained("facebook/regnet-y-10b-seer")
   model = RegNetModel.from_pretrained("facebook/regnet-y-10b-seer")
   

3. Extract Features and Infer: Process the image and infer using the model.

   inputs = feature_extractor(image, return_tensors="pt")
   
   with torch.no_grad():
       outputs = model(**inputs)
   
   last_hidden_states = outputs.last_hidden_state
   

4. Output Dimensions: The resulting tensor's shape is [1, 1088, 7, 7].

Suggested Cloud GPUs

For efficient execution, consider using cloud-based GPU services like AWS EC2 with GPU support, Google Cloud Platform, or Azure. These services offer scalable resources to handle the model's computational demands.

License

The model is released under the Apache 2.0 License, allowing for both personal and commercial use. For more details, refer to the Apache 2.0 License.